A Novel Feature Selection Method Using Whale Optimization Algorithm and Genetic Operators for Intrusion Detection System in Wireless Mesh Network

Abstract

Machine learning–based intrusion detection system (IDS) is an important requirement for securing data traffic in wireless mesh networks. The noisy and redundant features of network data tend to degrade the performance of the attack detection classifiers. Therefore , the selection of informative features plays a vital role in the enhancement to the IDS. In this paper, we propose a wrapper-based approach using the modified whale optimization algorithm (WOA). One drawback of WOA is that premature convergence results in a local optimal solution. To overcome this limitation, we proposed a method in which the genetic algorithm operators were combined with the WOA. The crossover operator was used to further improve the search space of whales, and the mutation operator helped to avoid being stuck in the local optimum. The proposed method selects the informative features in the network data, which helps to accurately detect intrusions. Using a support vector machine (SVM), we identified the types of intrusions based on the selected features. The performance of the improved method was analyzed by using the CICIDS2017 and ADFA-LD standard datasets. Our proposed method had better attack detection rate than the standard WOA and other evolutionary algorithms; it also had good accuracy and was suitable for IDS in the wireless mesh networks. The performance of the IDS was increased by selecting the informative features with the improved whale optimization algorithm. The attack detection ratio was higher than that of the standard WOA.